So Bitcoin's scalability relies on a mixture of writing new software when it's needed ("The core BitCoin network can scale to very high transaction rates assuming a distributed version of the node software is built. This would not be very complicated.") and in future switching to a two-tiered structure with "supernodes" transmitting data between themselves at a rate of >1GB/s. While I suppose that it's good that someone is thinking about this, it does not inspire confidence in bitcoin's ability to scale - the fact that it's theoretically possibly to write software allowing it to scale is very different from someone going off, writing this software, and demonstrating that it does scale.
The supernode architecture is mentally straightforward and has been the solution used by a number of previous P2P networks - eDonkey, Gnutella, and Skype come to mind as using supernodes or variants thereof. If you want to seriously criticize Bitcoin, one ought to have more than FUD and 'you haven't proven the opposite!'
